Specialists can improve aneurysm detection rates by using a deep learning algorithm that provides a second assessment of images already interpreted by radiologists, according to new findings published in Radiology. The algorithm was developed and tested using time-of-flight (TOF) MR angiography images due to the technique’s high sensitivity for detecting cerebral aneurysms.
“To our knowledge, this is the first study to devise and test an algorithm using TOF MR angiography images from multiple institutions under a variety of different conditions,” wrote Daiju Ueda from Osaka City University Graduate School of Medicine in Osaka, Japan, and colleagues. “A highly versatile algorithm such as this is desirable because the appearance of aneurysms differs depending on the type of imaging device used and the conditions under which the TOF MR angiography was performed. Additionally, there is variability in the diagnoses between individual radiologists.”
The authors collected MR images containing aneurysms from November 2006 to October 2017, separating them into a training data set, an internal test data set and an external data set. The algorithm was developed using the training data set, and the test data sets were used to test its effectiveness. Two radiologists independently performed a blind interpretation of any possible aneurysms detected by the algorithm.
Overall, the algorithm’s sensitivity was 91 percent for the internal test data set and 93 percent for the external data set. The algorithm improved aneurysm detection by 4.8 percent in the internal data set and 13 percent in the external data set. It did have a relatively low specificity, the authors added, because “the focus was on developing a supportive algorithm” that wouldn’t miss aneurysms.
“By providing a second assessment of the images, our algorithm could not only help radiologists to detect cerebral aneurysms, but also reduce the risk of overlooking aneurysms,” the authors wrote.